THE JOURNAL OF CHINA UNIVERSITIES OF POSTS AND TELECOMMUNICATIONS Volume 5, Issue, March 2008 SHENG Yu, PENG Mu-ge, WANG We-bo A ovel adaptive modulatio ad codig strategy based o partial feedback for ehaced MBMS etwork CLC umber TN929.5 Documet A Article ID 005-8885 (2008) 0-0048-07 Abstract The differece i lik coditio of broadcast/ multicast users ad the limitatio of uplik resource, make it difficult to utilize adaptive modulatio ad codig (AMC) i the ehaced multimedia broadcast ad multicast service (E-MBMS) etwork. To obtai the improvemet of system throughput, this study proposes a adaptive modulatio ad codig scheme based o partial feedback, by which oly partial users whose chael qualities are lower tha the system threshold eed to make a respose to the modulatio codig scheme () adaptatio procedure. By this ivestigatio, a adaptive scheme ca be itroduced i the E-MBMS etwork. Both the theoretical aalysis ad simulatio results demostrate the efficiecy of the proposed strategy, i which the performace is close to the ideal oe ad has a sigificat throughput improvemet whe compared with that of the fixed trasmissio scheme. Keywords E-MBMS, partial feedback, AMC, system throughput Itroductio Multimedia broadcast ad multicast service (MBMS) is a importat part of the uiversal mobile telecommuicatios system (UMTS) terrestrial radio access etwork (UTRAN) evolutio []. The MBMS feature, itroduced by the Third geeratio partership project (3GPP) i release 6 is proposed to use etwork ad radio resources more efficietly for trasmissio of multimedia services, both i the core etwork ad more importatly i the air iterface of UTRAN, where the bottleeck is placed to a large group of users. The physical layer i 3GPP log term evolutio (LTE) ca improve the trasmissio capacity sigificatly. To reduce the complexity of the system, the MBMS architecture i E-UTRA/E-UTRAN has bee modified to the ehaced MBMS, i Refs. [2, 3]. I Received data: 2007-06-2 SHENG Yu ( ), PENG Mu-ge, WANG We-bo School of Telecommuicatios Egieerig, Beijig Uiversity of Posts ad Telecommuicatios, Beijig 00876, Chia E-mail: godshy@gmail.com the ehaced MBMS (E-MBMS), the higher spectrumefficiet trasmissio scheme should be exploited. Adaptive modulatio ad codig is a good scheme for improvemet of spectrum-efficiecy [4 7]. I additio, for the orthogoal frequecy divisio multiplexig (OFDM) system, adaptive algorithms are also proposed i Ref. [8]. For the uicast system, lik adaptatio is possible because the chael status iformatio (CSI) ca be reported to the base-statio (BS) by the termial. However, for the broadcast/multicast system, the adaptatio mechaism ivolves a etwork respose based o feedback from more tha oe termial, ad the etwork respose caot be precisely tailored to the feedback. Geerally, this is because of the fact that the etwork receives a wide rage of feedback from differet termials ad more collisio would happe betwee termials if the uplik resource is limited. Cosequetly, the traditioal AMC scheme is difficult to utilize i a broadcast/multicast system. The coutig procedure is defied by 3GPP, to estimate the umber of MBMS users i target cells. I this article, the authors have proposed a ovel AMC strategy based o partial feedback mechaism for E-MBMS, which is similar to a coutig method ad oly eeds partial termials whose chael quality is lower tha the system threshold s respose to the procedure. Thus, the system will chage the level accordig to the partial feedback iformatio, to utilize lik adaptatio. By the authors ivestigatio, the uplik resource ca be saved ad oly a slight collisio would happe betwee termials. Therefore, adaptatio ca be used i the E-MBMS architecture whe the uplik issue is resolved, as the total throughout ca also be improved whe compared with the fixed mechaism. The rest of this article is orgaized as follows. Sectio 2 describes the proposed AMC strategy based o partial feedback for the E-MBMS etwork. Sectio 3 aalyzes the theoretical throughput of the statioary state i flat chael coditios. Simulatio results ad performace comparisos are doe i Sect. 4. The article is cocluded i Sect. 5.
No. SHENG Yu, et al.: A ovel adaptive modulatio ad codig strategy based o partial feedback for 49 2 AMC strategy based o partial feedback for E-MBMS 2. Trasmissio modes for E-MBMS For E-MBMS, there are two types of trasmissio modes described i Ref. [2]: Sigle-cell trasmissio ad multi-cell trasmissio. Sigle-cell trasmissio is where oe cell trasmits MBMS data oly to local campig termials, ad the tight, sychroous character betwee eighbor cells is ot required. Ulike sigle-cell trasmissio, multi-cell trasmissio requires that multicells trasmit the same cotets o the same frequecy badwidth simultaeously such as i the sigle frequecy etwork (SFN). The sigals from multicells will combie i air, ad uder this trasmissio mode, the tight, sychroous character is required. adaptatio could be applied for both sigle-cell ad multi-cell trasmissio. However, it should be more useful for sigle-cell MBMS. This is because, i the case of the multicells, co-ordiatio of the withi the SFN area is required. Moreover, because of the absece of performace beefits from multicells combiig, the sigle-cell MBMS service is expected to have a higher packet error rate tha the multi-cell MBMS service, especially at the cell edge. 2.2 Coutig-like partial feedback implemetatio It is difficult for broadcast/multicast users to retur all their chael quality iformatio. Therefore, the method which is similar to the coutig procedure i R6 MBMS is used i this ivestigatio. MBMS coutig is used to determie the imum trasmissio mechaism (poit to poit/poit to multipoit) for a give MBMS service []. Users who receive this service at the same time are called a group. The i the broad sese, it seems to trasmit the curret service i a poit to poit mode. Ad the proposed strategy is applied to oe give service ad its receivig group. As it is desirable to avoid brigig a large umber of termials for coutig purposes at the same time whe the uplik resource is limited, some radio resource maagemet (RRM) procedures may be take, such as, settig a access probability factor (PF). The the MBMS users decide whether to sed the uplik respose to the BS or ot, accordig to the PF value. O the basis of the coutig-like method, the partial feedback fuctio ca be implemeted. At the BS side, oe iitial NACK probability factor P NACK, which makes some users whose receive quality is lower tha the target threshold, respod to the AMC procedure, which is set ad set to all MBMS users through MBMS cotrol chael (MCCH), similar to Ref. [9]. M UL_idle PNACK = α () Nest Here, α is the ooverload factor, which is used to reduce the probability of vast users, to make resposes at the same time, geerally 0< α. Ad M UL_idle is the umber of available uplik chaels i the curret system. N est is a estimatio value for the umber of total users who are receivig the curret MBMS service. It is assumed that if the umber of users who respod to this procedure is NACK, the the system ca estimate the umber of the users who are ot satisfied with the curret level as N, accordig to the umber of reported users NACK NACK ad the PF value as P NACK. NACK NNACK = (2) PNACK For the accuracy of estimatio, the subprocedures should be k take ito accout ad the PNACK for the kth subprocedure is set accordig to the last N NACK as follows: MUL_idle M k UL_idle k PNACK = α = α P k k NACK (3) NNACK NACK The detailed coutig procedure should be refereced to Ref. []. Accordig to it, the uplik collisio ca be avoided if the system ca get the iformatio about the approximate umber of users who are ot satisfied with the curret level. The system model is show as Fig.. Comparig it with the traditioal uicast system, it is differet, i that, there is ot oly oe user to receive the sigal comig from the trasmitter. Whe the MBMS service trasmissio starts, the user would decide whether to sed the failig receivig otificatio NACK to BS or ot, accordig to the P NACK, which is set by BS o MCCH. 2.3 determiatio mechaism for E-MBMS The procedure of determiatio i this article is described i Fig. 2. Before the MBMS service trasmissio starts, the etwork side iitializes the system parameters ad seds them to the user equipmets (UEs). The iitial level is also set. After the radio access etwork (RAN) resource is set up, the UE should decide o the sigal to iterferece ad oise ratio (SINR) ad perform the probability test if its receivig SINR is lower tha the threshold. The, NACK is set to the etwork by UEs who pass this test. Fially, the level is adjusted accordig to the detectio of egativeackowledgmet (NACK).
50 The Joural of CHUPT 2008 Fig. System model Fig. 2 Proposed adaptive procedure The detailed flow is show i Fig. 3. Thresholds of SINRs are itroduced to decide if the users ca get to their satisfied status or ot, i the relevat levels. The iitial NACK probability factor P NACK is also set, ad for avoidig fluctuatio, a settig step legth Δ is ecessary here, to make the rate of the chage i moderatio. K sub is the maximal subprocedure umber. Durig the MBMS trasmissio, the user should estimate the effective SINR ad compare it with the threshold. It is clear that the user caot receive data packets correctly ad should ot be satisfied with the curret level whe the SINR is below the threshold. Therefore, they perform a probability test accordig to P NACK, to decide if they should respod to the AMC procedure or ot, by sedig NACK iformatio. Here, the iitial N est may ot be precise ad geerally it is set over-large. Cosequetly, there would ot be ay user to make a respose to the etwork ad users should wait for the ext probability iteratio whe the PF is set larger, just like i subcoutig. It ca be cosidered that all the users i this MBMS service group are satisfied with the curret level whe the value of PF is set at oe. Subsequetly, after the K sub subprocedures are complete, BS ca get the umber of users who have respoded to the curret AMC procedure. The etwork will set a less aggressive level i the ext data frame whe it detects the NACK otificatio set by some user. Otherwise, if the etwork does ot receive ay respose whe the highest probability factor (typically PF = ) is used, it should switch to a more aggressive level. Here, Δ, which stads for the step legth i the adjustmet procedure, is set, to avoid alterig too fast for levels ad the authors set 0< Δ <.
No. SHENG Yu, et al.: A ovel adaptive modulatio ad codig strategy based o partial feedback for 5 Fig. 3 Detailed flow of determiatio 3 Statioary state performace aalysis I this sectio, the authors aalyze the performace of the proposed strategy accordig to Ref. [0], whe it comes ito a statioary state. Actually, there are fiite modes to be selected i this system, ad the authors let the set Ω =,2,...,M stad for them. Here, it is assumed that the { } umber of available levels is M. The level of the ith subframe is deoted as L () i Ω. I this study, a real variable xi( xi R) is defied, to adjust the level with a fie step legth Δ, ad for simplificatio of aalysis, a fuctio F(): x L, ( x R, L Ω ) is also eeded, which ca covert x ito L. Thus, the defiitio ca be writte as follows: ; xi < L( i) = F( xi) = [ xi]; other (4) M; xi M + where [] deotes gettig the iteger of the variable. Accordig to the earlier-metioed algorithm, the level ca be adjusted usig a deduced method, accordig to the umber of reported users NACK. xi + ηδ ; NACK = 0 xi+ = xi NACKΔ ; NACK > 0, xi NACKΔ L ( i) (5) max{, L( i) }; NACK > 0, xi NACKΔ < L ( i) I Eq. (5), the effective steps are ηδ ad NACK Δ i the up-forward directio ad dow-forward directio respectively. Therefore, i the proposed strategy, the value of Δ should ot be over-high. η is used to balace the speed of adjustmet betwee the up-forward directio ad the dow-forward directio. The level L ( i+) of the (i + ) th data frame should be decided based o fuctio F( x i + ), ad x i+ ca be attaied accordig to Eq. (5). I this ivestigatio, the authors focus o aalyzig the throughput of the proposed strategy, comparig it with that of the ideal oe. Here, the term ideal deotes the chael
52 The Joural of CHUPT 2008 iformatio of all broadcast/multicast users kow by the etwork, whereas, it is difficult to realize i the curret system. L is used to express L = ( Ω ) ad for the idealizatio pilot SINR assistace method with exact feedback, oce the level ( L ) is assiged to the regio [ T ], T, whe oe user's pilot measuremet report CSI falls ito regio [ T ], T, the user should otify BS, ad trasmit the data with L. Here, T is the threshold of the SINR regios betwee L ad L, ad T is the threshold betwee L ad L +. I this sectio, oly the differece of throughput performace is cosidered betwee the proposed strategy ad the ideal AMC mechaism. For each L Ω, there is a relevat throughput (bit/symbol). It is assumed that the SINR of the curret frame, which is received by the termial, is χ, ad for the ideal scheme, the value of χ is kow by BS. Accordigly, the imal level for the curret frame ca be selected as follows: L = L; T < χ < T (6) Here, i the broadcast/multicast case, χ ca be set to the mea SINR received by users who are receivig the same MBMS service, ad the the L is assumed to be the imal level for this group of users. Ad the, the system throughput is γ = γ, which is i the curret level L. Here, the authors do ot cosider retrasmissio of the data packet whe it could ot be decoded correctly. I the proposed strategy, BS caot select the imal level immediately ad it eeds several steps to get to L. ii L is deoted as the iitial level before the MBMS service trasmissio starts ad there are two coditios to be discussed: ii If L < L The BS should icrease the level accordig to settig the step legth ad the directio is ii L : from L L ascedig It is assumed that the step legth is Δ, ad the level should get to L i at the most S a steps. Accordig to Eqs. (4) ad (5), S a ca be deoted as follows: ii L L Sa = (7) ηδ It should be oted that the real variable x eed ot to be chaged whe the probability factor P NACK is icreased to the maximal value, to simplify the aalysis. Therefore, oly oe step is cosidered after P NACK is set to oe. ii If L > L γ With the same deductio as i coditio (a), BS should decrease the level ad the directio is ii L :from L L descedig The it should get to L i at the most S d steps. Ad it ca be deoted that 0.5 ii Sd = + ( L L ) (8) NACKΔ Accordig to Eq. (5), it ca be see that it is differet with the ideal scheme, where the level status of the proposed strategy should trasit two modes periodically betwee L ad L +, whe it comes ito covergece as show as Fig. 4. I Fig. 4, ρ = / ηδ ad a simple coclusio ca be draw that the period of statioary state is + ρ. Cosequetly, the system throughput ca be expressed as follows: γ ηδ ρ γ = = γ (9) + + ρ ηδ Fig. 4 State trasfer i adaptatio process From Eq. (9), it ca be see that the system throughput of the proposed strategy is close to γ, which is the throughput of the ideal oe whe Δ is reduced. However, the value of Δ caot be set over-low because of the pace of covergece. Tradeoff should be cosidered durig simulatio. 4 Simulatio ad aalysis It is assumed that there are four differet modulatios available ad the codig status i the simulatio system is as show i Table. Table Modulatio ad codig rate Modulatio type Turbo codig rate level QPSK /2 level 2 QPSK 3/4 level 3 6QAM /2 level 4 6QAM 3/4 Simulatio is performed uder the assumptios as follows: the multipath chael is a Rayleigh chael with the expoetial fadig gai profile. The umber of subcarriers is 28 ad the received SINR ca be calculated accordig to Eq. (0). 2 SINR = α SINR arg (0)
No. SHENG Yu, et al.: A ovel adaptive modulatio ad codig strategy based o partial feedback for 53 Here, α is the chael fadig value ad SINR avg is the predefied SINR. A summary of the simulatio parameters is give i Table 2 a little later i the text. Here, because of poit-to-multipoit trasmissio, the imal level caot be set accordig to some user s SINR. I MBMS, all users who receive the same service ca be regarded as a whole group ad it is assumed that all SINRs of this user-group are kow by BS. The SINR arg is set accordig to all users SINRs i this group. Therefore, the imal level i this study, for the curret MBMS service trasmissio ca be decided accordig to SINR arg. The system throughput (bits/symbol) of the ideal scheme ad proposed strategy are compared i this ivestigatio as show as Fig. 5 ad Fig. 6. Table 2 Simulatio parameters ad assumptios Parameters Values Propagatio coditios 3-Path rayleigh Vehicle speed 3 km/h Subcarrier umber 28 Chael estimatio Ideal chael estimatio Available levels /2 QPSK, 3/4 QPSK; /2 6 QAM, 3/4 64 QAM update period frame cotrol delay 0 HARQ No Simulatio time 5 000 packets per SINR value Number of MBMS users 0 300 Δ 0., 0.2, 0.4 Figure 5 shows the compariso of system throughput for six coditios, which are, the ideal oe, proposed strategy, ad four fixed modes. Because of multiusers receivig oe MBMS service at the same time, the system throughput is replaced by mea throughput of all the group users. The adaptive scheme, based o partial feedback mechaism, ca track the receipt quality of the group users who are receivig the curret MBMS service. Cosequetly, the performace of the proposed oe is better tha that of the fixed modes. However, ot all the users SINRs are kow to the BS, hece, the performace of this scheme ca oly be close to the ideal oe, but ot get better tha it. Figure 6 shows the differece of the proposed strategy performace whe the step legth Δ is chaged. It ca be see that the performace of Δ = 0. is closest to the ideal oe ad whe Δ = 0.4, the performace of the proposed method begis to fall. The collisio probability of partial feedback mechaism based o coutig-like method i MBMS architecture is also preseted i this ivestigatio, as show as Fig. 7. For the broadcast/multicast system, the adaptatio mechaism ivolves a etwork respose, based o feedback from more tha oe termial. By itroductio of partial feedback, the collisio ca be reduced efficietly whe the resource from the uplik is limited. I Fig. 7, it is assumed that the umber of available uplik chaels is 5, 8, ad 0 respectively. The collisio probability icreases obviously whe the uplik resource is i scarcity, such as, M UL_idle =5, which stads for the umber of available uplik chaels. Whe gets M UL_idle to 0, the probability falls to about 0% ad it ca also decrease it by settig a ooverload factor α i Eq. (). Fig. 5 System throughput for differet schemes Fig. 7 Collisio probability for differet uplik resource Fig. 6 Diversificatio of Δ ad correspodig cases Figure 8 shows the probability distributio fuctio (PDF) of the subprocedure i oe AMC respose. From this figure, it is foud that more tha 90% of the AMC respose procedures ca be completed i the first two subprocedure steps. It is helpful to reduce the time delay of the partial feedback procedure i the actual system.
54 The Joural of CHUPT 2008 5 Coclusios Fig. 8 PDF of Subprocedures For broadcast/multicast system, it is difficult to utilize the AMC mechaism because the adaptatio procedure ivolves vast etwork resposes from more tha oe termial ad the chael status caot be precisely tracked. I this study, a ovel AMC strategy based o partial feedback is preseted to the E-MBMS etwork, by which, oly partial users whose chael qualities are lower tha the system threshold eed to respod to the procedure. The collisio probability is reduced efficietly by coutig-like partial feedback whe the uplik resource is limited ad accordig to the theoretical aalysis ad simulatio results, the performace of the proposed strategy is close to the ideal oe, as the system throughput is sigificatly improved compared to that of the fixed E-MBMS trasmissio, without adaptive mechaism. Ackowledgemets This work is supported by the Natioal Natural Sciece Foudatio of Chia (6057220), the Hi-Tech Research ad Developmet Program of Chia (2006AA0Z257). Refereces. Itroductio of the multimedia broadcast multicast service (MBMS) i the radio access etwork (RAN)- Stage 2. 3GPP TS 25.346 V7.2.0, 2006 2. Evolved uiversal terrestrial radio access (E-UTRA) ad evolved uiversal terrestrial radio access etwork (E-UTRAN). 3GPP TR 25.83 V7..0, 2006 3. Physical layer aspects for evolved uiversal terrestrial radio access (UTRA). 3GPP TR 25.84 V7..0, 2006 4. Goldsmith A J, Chua S G. Adaptive coded modulatio for fadig chaels. IEEE Trasactios o Commuicatios, 998, 46(5): 595 602 5. Zhou S, Georgios B G. Adaptive modulatio for multi-atea trasmissios with chael mea feedback. IEEE Trasactios o Wireless Commuicatios, 2004, 3(5): 626 636 6. Alouii M S, Goldsmith A J. Adaptive modulatio over Nakagami fadig chaels. Kluwer Joural o Wireless Commuicatios, 2000, 3( 2): 9 43 7. Fa Che, Che Mei-ya, Su Li-ju, et al. Research o threshold adjustmet algorithm i adaptive modulatio ad codig. The Joural of Chia Uiversities of Posts ad Telecommuicatios, 2006, 3(2): 5 9 8. Wog C Y, Cheg R S, Be Letaief K, et al. Multiuser OFDM with adaptive subcarrier, bit ad power allocatio. IEEE Joural o Select Areas i Commuicatios, 999, 7(0): 747 758 9. Sheg Yu, Peg Mu-ge, Wag We-bo. Adaptive traffic multiplexig for MBMS i the UMTS system based o the scalable codig. Proceedigs of st Iteratioal Coferece o Commuicatios ad Networkig i Chia (CHINACOM 2006), Oct 25 27, 2006, Beijig, Chia. Piscataway, NJ, USA: IEEE Computer Society, 2006: 5p 0. Yag Hog-we, XiaoXua, Guo We-bi, et al. A AMC/ HARQ scheme based o ACK feedback. Joural of Beijig Uiversity of Posts ad Telecommuicatios, 2004, 27(2): 29 33 (i Chiese) Biographies: SHENG Yu, male, Beijig Uiversity of Posts ad Telecommuicatios, Ph. D. Cadidate, iterested i the research o wireless broadcastig ad resource schedulig algorithm. PENG Mu-ge, Beijig Uiversity of Posts ad Telecommuicatios, Ph. D., istructor, iterested i the research o wireless commuicatio 3G/B3G system ad radio resource schedulig. WANG We-bo, Beijig Uiversity of Posts ad Telecommuicatios, professor ad doctor supervisor, iterested i the research o sigal processig, mobile commuicatios ad wireless Ad-hoc etwork.